Lung Cancer Nodules Detection via an Adaptive Boosting Algorithm Based on Self-Normalized Multiview Convolutional Neural Network

Figure 3

Schematic of the proposed AdaBoost-SNMV-CNN. In the preprocessing stage, lung cancer image data were divided into training (80%) and testing (20%) subsets. To train the AdaBoost-SNMV-CNN, the data weight was initialized. The first SNMV-CNN was trained using the initial data weight, then SNMV-CNN was used to update the data weight of the second SNMV-CNN. This process continues until the SNMV-CNN was trained. Finally, the SNMV-CNN predictions for the testing subsets were combined using a weighted voting approach to get results (such as nodules or nonnodule).